The effect of discretionary snack consumption on overall energy intake, weight status, and diet quality: A systematic review

The consumption frequency and portion size of discretionary snacks are thought to contribute to a greater food intake and risk of overweight or obesity in the developed world but evidence from epidemiological studies is inconclusive. To investigate this, we systematically evaluated evidence on the effects of discretionary snack consumption on weight status, energy intake, and diet quality. Articles involving discretionary snacks reported against the outcome measures of any primary, peer‐reviewed study using human participants from free‐living conditions for all age groups were included. A total of 14,780 titles were identified and 40 eligible publications were identified. Three key outcomes were reported: weight status (n = 35), energy intake (n = 11), and diet quality (n = 3). Increased discretionary snack consumption may contribute modestly to energy intake, however, there is a lack of consistent associations with increased weight/BMI. Although cross‐sectional analyses offered conflicting findings, longitudinal studies in adults showed a consistent positive relationship between discretionary snack intake and increasing weight or body mass index. Given that experimental findings suggest reducing the size of discretionary snacks could lead to decreased consumption and subsequent energy intake, food policy makers and manufacturers may find it valuable to consider altering the portion and/or packaging size of discretionary snacks.

mean BMI, during the COVID-19 pandemic in both adults 3 and children. 4 Accordingly, the increase in individuals living with overweight and obesity has coincided with an increase in obesity-related comorbidities such as cardiovascular disease (CVD), type 2 diabetes mellitus (DM), and cancer 5 as well as overall reduced quality of life 6 and more recently poorer outcomes following COVID-19 infection. 7On top of the human health cost, the financial burden of obesity, in terms of health care expenditure, has been calculated to amount to USD 190 billion in the United States alone in 2012, 8 with further costs associated with lost productivity/absenteeism due to excess weight. 8though the development of excess weight is often oversimplified to result from an energy imbalance, derived from excess calorie intake in relation to insufficient calorie expenditure, the reality of its development is more complex.[11][12][13] Increased food portion size (FPS) and frequency of consumption are factors that are believed to play a role in this increased propensity to consume excess calories, 14 which may result in increases in weight status.Additionally, so is the increased availability of, readily available and hyperpalatable snack foods. 11Furthermore, diets that contain a large proportion of such nutrient-poor snack foods may reduce overall diet quality (DQ) and contribute to poorer health status. 15e greater frequency of consumption and larger portions of snack foods may partly contribute to increased overall food intake and risk of individuals becoming classified as living with overweight or obesity, [16][17][18] although this effect may vary depending on the population studied and it is noteworthy that results from observational research are inconsistent. 17,19These inconsistent results may be influenced by variations in the definition of snacks used in research.Such variation in definition is potentially a major obstacle to the standardization of research relating to the effects of snack intake on diet-or weightrelated outcomes. 20,21Depending on the specific study, the definition of snack can vary from being based on calorie content 22 to time of day consumed (i.e., items usually consumed apart from main meals) 22,23 and even foods self-defined by study participants as snacks. 24umpakari et al 25 proposed defining foods as core (formed of the five food groups of fruit, vegetables, cereals, meat and alternatives, and milk and alternatives) and non-core (all other foods) in relation to their desirability within a diet.Nutrient cut-offs (e.g., high in sugar or fat) may also be used to classify foods; however, this does not distinguish between core and non-core foods.For example, almonds (30 g) which are associated with health benefits 26,27 contain approximately 50% more fat than a packet of crisps (30 g). 28Therefore foods defined as discretionary snacks (DS), can be considered as non-core foods usually consumed outside of main meals, that typical dietary guidance (for example the UK Eatwell Guide 29 ), does not recommend for regular consumption (e.g., crisps, chocolate, cakes, and confectionery). 25The inclusion of sugar-sweetened beverages (SSBs)   in the definition of discretionary snacks may lead to considerable differences in outcomes compared with definitions that focus solely on solid foods.The reasoning for this is that SSBs alone are known to be the largest single dietary source of added sugars in countries such as the United States 30 and their consumption may disproportionately relate to body mass gain due to reduced satiety and an incomplete compensatory reduction in energy intake (EI) compared to solid foods. 31Therefore, within this systematic review, SSBs were not included in the definition of DSs.
The objective of this systematic review was to assess the effects of DS consumption, on weight status, EI, and DQ, defining DS as non-core foods that typical dietary guidance does not recommend for regular consumption and that are usually consumed outside of main meals.

| METHOD 2.1 | Search strategy
The review followed the PRISMA 32 guidelines (Supplement 1) and was registered in the PROSPERO database (ID CRD42021295446).
A systematic search was conducted using six scientific databases: PubMed, Psych Info, Cochrane, ERIC, CINAHL, and Medline in December 2021.Keywords used in the search were drawn from previous relevant literature.The search strategy was organized around DS, defined as non-core foods that typical dietary guidance does not recommend for regular consumption and that are usually consumed outside of main meals (adapted from Gage et al 33 ) and three outcomes: dietary EI (any measure), weight status (any measure e.g., BMI, percentage body fat [%BF] or waist circumference), and DQ (measured via validated indices).Limits were set to include only journal articles published in the English language.Further titles were identified by cross-referencing from these sources.

| Eligibility criteria
Articles involving DS reported against the outcome measures of any primary, peer-reviewed study using human participants within noninstitutionalized living conditions for all age groups were included.
Studies that did not report a relevant outcome or were validation papers were excluded.Likewise, studies focused on participants with health conditions or using drugs or supplements that may affect appetite, and participants living in countries with a human development index of <0.8, as published in the latest Human Development Report, 34 were also excluded.After the removal of duplicates, articles were screened independently by three reviewers (H.G., D.M., and R.K.) against the eligibility criteria, using the online tool Covidence. 35

| Data extraction
Data were extracted from each publication by two of three reviewers (H.G., D.M., and R.K.) and consensus agreed with the third author.Information was extracted on study characteristics, methods, population, interventions, and comparisons as well as all available records matching our a priori selected outcome measures.Corresponding authors were contacted for missing or additional information when necessary.

| Quality assessment (QA)
Eligible papers underwent a QA using either the checklist for quality assessment of controlled intervention studies or the checklist for quality assessment for observational cohort and cross-sectional studies, as appropriate. 36These QA tools were designed by the US National Heart Lung and Blood Institute to assist reviewers in focusing on factors essential for critical appraisal of the internal validity of a study.The tools include items to evaluate potential flaws in study methods or implementation.QA was completed for each publication by two of three reviewers (H.G., D.M., and R.K.) with any discrepancies discussed until a consensus was reached (Supplements 2 and 3).

| Data synthesis
Due to the heterogeneity in exposure metrics and methodologies used across eligible studies, a meta-analysis was not possible.Data are presented in a narrative format, using the Synthesis Without Meta-analysis (SWiM) guidelines. 37

| RESULTS
A total of 14,780 titles were identified.After removing duplicates, 7875 records were retained.Of these, 7483 studies were excluded after reviewing titles and abstracts and 392 full-text articles were assessed for eligibility, resulting in a final sample of 31 eligible publications.A manual review of the references of these publications led to the addition of a further nine articles, and a final sample of 40 publications (Supplement 1).

| Study characteristics
The characteristics of the included studies are presented in Table 1, with data from studies outlined in Tables 2, 3, 4, 5, and 6.The included studies were mainly cross-sectional in nature (n = 27). 15,18, Addi1][72][73][74][75] Both male and female participants of varying cultural and socioeconomic backgrounds, were included.9,62-66,70,72-75 One study, O'Neil et al 68 tracked behavior from childhood through to adulthood.68][69][70][73][74][75] Sixteen studies reported on participants living in European countries, 18,[38][39][40]45,48,[50][51][52][53]59,62,66,67,71,72 whereas two studies were reported in Australia, 60,61 two in Canada, 42,55 and one in Saudi Arabia.43 Different studies reported data using a range of different foods as a measure of DS (Supplement 4). Usng the QA tools, 36 the quality ratings for all 40 studies were fair to good, although substantial heterogeneity was observed.Strengths of the studies included clearly identified research questions, description and definition of study populations, and clear definition and measurement of relevant variables.Weaknesses in some studies included failure to assess the impact of DS intake over a timeframe sufficient to reasonably assess an association.Furthermore, some studies did not examine different levels of DS intake as related to the outcome. The results from the QA are presented in Supments 5 and 6.

| Outcome measures
Three key outcomes reported within the included literature were
Similarly, in a large study of English adults, O'Connor et al 62 identified that individuals living with a BMI > 25 kg/m 2 reported consuming statistically significant greater amounts of crisps, chocolate, ice cream, and sweets (g/10 MJ/day) but not cakes and biscuits compared with those living with a BMI < 25 kg/m 2 .Rippin et al 59 examined how FPS might vary with BMI in French and UK adults, however, only cakes in the French group had a significant association between FPS and BMI, where FPS increased with each BMI point increase.
Conversely, two studies 47,57 identified a negative association between chocolate intake and weight status.Golomb et al 47 reported that an increased frequency of chocolate consumption was linked to lower BMI, even after multiple adjustment models, despite also being linked with greater calorie and saturated fat intake.Using national data from American adults, O'Neill et al 57 reported that candy consumers had lower weight and waist circumference than non-candy consumers; in addition, total and sugar candy consumers also had a lower mean BMI than non-candy consumers, although a lower mean BMI was not seen in chocolate candy consumers.Further to these two studies, Just and Wansink 49 reported an inverse relationship between BMI and sweet and salty DSs; however, when excluding the most extreme BMI classifications as outliers, no relationship was found.Additionally, although Matsumoto et al 54 reported a significant change in BMI with different chocolate consumption frequencies, the clinical relevance was minimal and the association did not follow a linear pattern.
Further to the findings of Just and Wansink 49 and Matsumoto and colleagues, 54 nine further studies identified no relationship between DS intake and weight status.Anyżewska et al, 40 in a sample of male Polish Army personnel, reported no association between chocolate, chocolate candies, and candy bars; non-chocolate candies; biscuits and cakes; ice cream and pudding; and salty snacks with BMI or a fat mass index.In a population of post-menopausal Polish women, G orna et al, 48 observed that the frequency of consumption of sweet or salty DSs, which ranged from less than twice per month to at least six times a day, was not associated with the development of overweight or obesity.Despite 95% of a study population of Canadian University students consuming DSs, Brunt et al 42 reported no relationship between sweet baked goods, salty snacks, candy, and BMI.Similar findings were reported in the United Kingdom, 50,59 between portion sizes (g) of food groups consumed with BMI status, adjusted for under-reporting, social class and physical activity, and American 56 cohorts, which reported the category of candy consumption, adjusted for sex, age, and race/ethnicity, despite being positively associated with daily EI. 56Although Barnes et al 15 reported a positive association between desserts and sweets, no association was reported between chips, crackers, ready-to-eat cereals, popcorn, and related products and BMI.Similar discrepancies were reported in children.Lioret   T A B L E 3 Cross-sectional studies reporting DSs and EI and/or weight status for adults.
Author   ice cream, chocolate spreads, and all savory appetizers and biscuits, only the portion size of biscuits and sweetened pastries was positively associated with children being classified as living with excess weight.
Likewise, Albar et al 38 reported a positive association between intakes of only buns, cakes and pastries, and biscuits with BMI, that is, for each 10 g of biscuits or cakes consumed, BMI increased by 0Á28 and 0Á19 kg/m 2 , respectively.Kosti et al 52 reported that eating sweet snacks was positively associated with overweight/obese status in male, but not female Greek adolescents.
Seven additional studies identified a negative association between snack intake and weight status. 39,41,43,45,55,58,61Andersen et al 39 reported that in a sample of Norwegian 8-13-year-olds, those consuming the highest quartile of sweet intakes had 50% lower odds of living with excess weight compared with those in the lowest quartile of sweet intake.In an Australian trial, Babajafari et al 61 reported an increase in the consumption of cakes/biscuits was associated with a decrease in the odds of living with excess weight among adolescents, although no association between sweets/lollies and BMI was reported.Similarly, adolescents from Saudi Arabia reported savory snacks were inversely associated with BMI for both males and females, whereas sweet DS were inversely associated with BMI in males. 43 a multinational sample of European adolescents (Greece, Germany, Belgium, Crete, France, Hungary, Italy, Sweden, Austria, and Spain), Cuenca-García et al 45  Bandini et al 41 compared the intake of high-calorie, lownutrient-dense foods among those living with and without obesity.
They observed that daily caloric intake from chips was similar between adolescent groups yet, caloric intake from candy, baked goods, and ice cream was significantly higher among adolescents living without obesity.However, after adjustment for under-reporting, only the intake of ice cream remained significantly higher.Mercille et al 55  Four further studies identified no relationship between DS intake and weight status. 44,51,52,60Couch et al 44     consumption of DSs.Kerr et al 51 reported there was no difference in chocolate confectionery or crisps and savory snack intake between adolescents with a healthy weight or with excess weight or obesity in either a UK or Northern Irish cohort.Furthermore, Schumacher et al 60 reported BMI was not associated with the consumption of packaged snacks, baked sweet products, and confectionery products in female adolescents from low-income communities.

Longitudinal
8][69] All the adult studies reported a positive association between their chosen DS food and increasing weight or BMI.In a large prospective cohort study of postmenopausal American women, Greenberg et al 64 reported that each 1 oz/day increase in chocolate-candy consumption was associated with a greater weight gain of 0.92 kg (0.80, 1.05), after 3 years.Similarly, in an earlier sample of female participants, Greenberg et al 63 reported more frequent chocolate intake was associated with greater prospective weight gain over 6 years, in a dose-response manner.Hendriksen et al 66 reported that after adjustment for potential confounders, a 100-kcal higher intake of savory DSs was significantly associated with an annual weight gain of 9.9 g/year (95% CI: 2.2, 17.5 g/year) in two of the three towns included in the study over an average follow-up of 8.1 years.In a shorter, weight loss intervention trial, Harris et al 65 reported that a mean decrease of 4.7 servings per week of sweets over 18 months was associated with a decrease in BMI of 0.12 kg/m 2 .

Cross-sectional
A positive relationship between DS consumption and EI was reported in cross-sectional studies in both adults (n = 3) and children (n = 2).Increased chocolate 47,54,57 and sugar candy 57 consumption was associated with greater calorie intake in adults.Golomb et al 47 reported that the frequency of chocolate consumption was linked to greater calorie and saturated fat intake among healthy men and women (20 to 85 years).Cross-sectional analysis from Matsumoto et al 54 examined the association of chocolate consumption and reported EI significantly increased by 335 kcal/day comparing no chocolate consumption to ≥2 servings a week.O'Neill et al 58 reported chocolate candy consumers had higher energy, total fat, saturated fatty acid, and added sugar intakes compared to nonconsumers, whereas sugar candy consumers had higher energy and added sugar intakes and lower total fat and saturated fatty acid intakes than non-consumers.
One study reported on cross-sectional EI in children, and the findings replicated those of adults. 58O'Neill et al 58 used NHANES data from American children (n = 11,181) and reported chocolate candy consumers had higher intakes of energy, total fat, saturated fatty acids, and added sugar than non-consumers.Sugar candy consumers had higher intakes of energy and added sugar and lower intakes of total fat and saturated fatty acids.

Experimental
4][75] Increased consumption of a range of DSs, including cookies, cake, sugar confectionery, chocolate confectionery, and savory snacks were reported to be positively associated with an increase in EI.Four RCT studies 70,[73][74][75] modified the package size of the DSs they provided to assess if this influenced the amount consumed.These studies were conducted over a range of time periods, from a week 74 (consuming three different snack brands chosen out of a possible 10, including crisps, crackers, pretzels, and biscuits) to 3 days 73 (potato chips, cheese crackers, cookies, and candy), and one eating episode 70 (crackers).Although Raynor and Wing 73 reported that doubling the amount of food provided was associated with an 80.7% increase in calories of food consumed there was no effect on package unit size.In contrast, both Stroebele et al 74

| DQ
Only three studies investigated the association between DS consumption and DQ 15,57,58 and have reported mixed findings.In adults, an increase in desserts and sweets (including cakes, cookies, pies, candy, sugar, and sweets), was significantly inversely associated with a Healthy Eating Index (HEI) score, 15 whereas sweet candy consumers had a significantly lower HEI score compared with non-consumers. 57 contrast, an increased intake of savory DSs (chips, crackers, readyto-eat cereals, and popcorn) 15 and chocolate confectionery had no significant association with HEI scores. 57In children, chocolate consumers reported a mean lower dietary quality (HEI) compared with non-consumers, although no difference was observed with sugar candy consumption. 58Regression analysis suggested that neither chocolate nor sugar candy consumption predicted DQ scores. 58

| DISCUSSION
This review examined studies that investigated the association between DS consumption and measures of weight status, EI, or DQ.Consistently, an increased intake of DSs was positively associated with EI.However, there was no consistent association between DS intake with increased weight/BMI.The lack of consistency is likely due to differences in research designs as cross-sectional studies, which are weaker designs not capable of establishing causal relationships, showed mixed findings, whereas longitudinal studies in adults, which provide more robust evidence, showed a positive relationship between DS and weight status.Similarly, research reported mixed results for the association between DS intake and DQ, with two studies reporting that sweets and desserts, but not chocolate, were associated with reduced DQ in adults and only one study, in children, finding that consumption of chocolate but not sweets was associated with decreased DQ.
Increased consumption of a range of DSs has been consistently reported to be positively associated with an increase in EI in both adults 47,54,57,[72][73][74][75] and children. 58,71In this review, a wide range of DS measures were identified; however, overall, the results suggested that using portion-controlled packaging can help reduce FPS and energy consumed.These findings support recent systematic reviews that, increased portion size is positively associated with EI in children 76 and adults. 77,78Reducing portion size, availability, and appeal of largersized portions, packages and tableware have the potential to reduce food consumption. 79A plausible explanation for this is that people may consume DSs when they are neither hungry nor genuinely satiated.The decision about the appropriate amount of food to consume is therefore not a response to a physiological requirement but may be based on food hedonics, that is, liking and wanting. 80It may be suggested that people interpret that consuming one unit of food is the appropriate amount to consume regardless of the size of the food items. 72though two studies reported higher EIs derived from a measure of DS consumption in individuals living with obesity compared to those without, 18,41 in studies investigating the effects of DS intake, portion size, and frequency, on various aspects of weight status, the results varied considerably.Among adults nine cross-sectional studies, 15,40,42,[48][49][50]56,57,59 indicated there was no relationship between a measure of DS intake and weight status, whereas five studies 15,18,46,59,62 reported a positive relationship, and four studies 47,49,54,57 indicated an inverse association. However,among the latter studies, the result was not maintained when excluding individuals with extreme weight categories 49 and was deemed not to be clinically significant in another.54 Among children, the findings continue to be inconsistent, with seven studies 39,41,43,45,55,58,61 reporting an inverse association; three studies, 38,52,53 a positive association; and 11 studies, 38,41,43,44,[51][52][53]55,60,61 no relationship between an individual's DS intake and weight status.
These conflicting findings could be at least partially explained by the physical activity levels of the participants as there is strong evidence demonstrating that higher physical activity levels can attenuate body mass gain. 81In adults, for example, Anyżewska et al 40 reported high physical activity levels which may have attenuated the impact of high DS consumption on BMI.Interestingly, O'Neill et al 58 reported that higher candy consumers had lower BMI levels but stated that physical activity did not influence the main findings but reduced the association with BMI.However, they did not report the adjusted data for physical activity levels, leaving it unclear how much of an attenuation it had on BMI.In contrast, Kelly et al 50 and Murphy et al 56 did adjust for physical activity levels and still observed no association between BMI categories and DS intake.In children, association of DS intake with weight status was not adjusted for physical activity levels in 5 of the studies 38,51,52,55,60 that showed no effect of DS intake.
Furthermore, cross-sectional studies only provide a snapshot of dietary intake and weight status making it difficult to determine causal effects. 82For example, reverse causation could also play a role in cross-sectional studies. 83It is possible that when data was collected those with higher obesity levels may have reduced DS intake to help reduce their obesity levels.In contrast, longitudinal studies are more robust to these limitations and offer better insights into DS intake and weight status.
All four longitudinal studies in adult populations reported a positive association between DS intake and weight status.Conversely, all three studies in children's populations reported inverse associations between DS consumption with weight status.We are unable to determine the reason for the greater proportion of studies reporting an inverse association in child and adolescent populations, compared with adults, a discrepancy that was most notable in longitudinal research studies.It might be speculated that children, due to their rapid growth and inherently faster metabolic rate, may consume more DS foods without negative impacts on their weight status. 84,85Furthermore, rapid growth during childhood often includes increases in both weight and height, limiting the use of BMI as a valid measure of adiposity in this age group.It should also be mentioned that BMIz is considered by some to be a poor predictor of both adiposity and changes in adiposity over time, in young children. 86,87This may be partially due to the greater risk of misclassifying rapidly growing children as living with overweight or obesity, by using age-stratified BMI or BMIz. 88Although difficult to compare due to differing definitions of snacks, previous systematic reviews have observed similar conflicting findings in adults 89 and children 90 with regards to the role of snacks contributing to obesity levels.Likewise, authors have recognized that the differing methodological approaches to addressing the role of snacks and weight status are a key challenge to producing clear conclusions.
Typically, diets of high quality are associated with a reduced risk of mortality and non-communicable diseases. 5For example, the risk of all-cause mortality, cardiovascular disease, cancer, type 2 DM and neurodegenerative disorders are reduced by 22%, 22%, 16%, 18%, and 15%, respectively, due to following a high-quality diet. 91erefore, diets that are high in nutrient-poor foods may contribute to adverse health outcomes due to low overall DQ.Previous research has indicated that snacking may be positively associated with higher DQ. 92,93However, specific snack foods may affect DQ differently due to their nutrient composition.For example, Barnes et al 15 observed that consumption of nuts, fruit, and fruit juice was positively associated however, desserts and sweets were negatively associated with DQ.
Further to findings reported by Barnes and colleagues, 15 DQ was reported to be lower in sugar candy consumers compared with nonconsumers; however, no difference was observed between chocolate consumers and non-consumers. 58The nutrient composition of snack foods may be a determinant of how it affects overall DQ.Sugar candies are mainly a source of sugar in crystalline or semisolid form with rarely any addition of products with nutritional properties; in contrast, chocolate candies contain sugar but also a mixture of processed cacao, and cocoa butter, with additional products, including milk, fruit, and nuts, among other ingredients, which may contribute slightly more positively to DQ.For example, dark chocolate is a rich source of phytonutrients and has previously been associated with an improved lipid profile leading to reduced CVD risk. 94,95However, in children, the trend was equivocal as chocolate consumption, compared to non-consumption, was associated with lower DQ, but sugar candy consumption was not. 58Despite this, neither chocolate nor sugar candy consumption predicted DQ. 58 Additional research is required to understand how DS foods contribute to DQ in adults and children.
In interpreting these results, consideration should also be given to the type of research employed in each study.Cross-sectional studies are inherently limited in using data from a single point in time and as such may not represent normal or ongoing behaviors or states. 82Longitudinal and experimental studies provide the advantage of following cohorts, which may differ by eating behavior, and observing changes, which may occur over a time course sufficient for changes in weight status to occur. 82As such, longitudinal and experimental studies are considered to offer more valid epidemiological insights, 82,96 and in this systematic review, longitudinal data relating to DS intake and weight status may offer more relevant information than similar but crosssectional research.Likewise, the different analysis techniques used to interpret the data, including what adjustments were included within the statistical model, may go some way to explaining inconsistencies in findings.

| Strengths and limitations
Within this review, the diversity of studies, including cross-sectional, longitudinal, and experimental, in both adult and child/adolescent populations, allows for both a general and more focused depiction of the current literature related to DS consumption.However, the diversity of individual study designs, exposure and outcome measures, study durations, and so forth adds a great deal of heterogeneity making a meta-analysis less reliable; hence, our decision to not include such an analysis.Indeed, the sheer diversity of definitions of DSs and/or different dietary assessment methods among the included studies and the literature in general was and is an issue of relevance.
No generally accepted definition exists of what constitutes a DS, and distinctions between how DS consumption is presented in the literature are often ambiguous.6][47][48][49]54,[57][58][59][60][61][63][64][65]74 Inconsistencies in the definition of DSs make comparisons between studies difficult to interpret.The study definition of DS used herein was chosen to solely focus on foods that add variety to the diet but are not recommended for consumption such as chocolates, crisps, and sweets.As such all studies that specifically focused on these types of DS were included in this review.However, as a result of this definition, some studies may have been excluded that grouped these DS with snacks that are typically recommended for consumption such as nuts, yogurts, and fruits.By excluding such snacks this allowed the study to reveal the association of DS without confounding the results with core foods that are beneficial for consumption.
A further concern of all such nutritional epidemiology research is the reliability of diet intake assessment and the possibility of underreporting.Recording dietary intake in population-level studies is inherently difficult with participants providing inaccurate records due to poor memory, inability to record food intake near to time of consumption, inaccurate estimation of portion sizes, and confusion regarding the definition of certain food categories, such as DSs. 97Furthermore, the misreporting and in particular, under-reporting of food intakes (either intentionally or unintentionally) may present considerable limitations to any conclusions drawn from such data, and appropriate adjustment during data analysis is recommended. 98,99However, such adjustment is not always carried out, as can be seen by the small number of studies included in this review which explicitly mention that the use of an adjustment for under-reporting was made in their analysis.
Finally, it should be emphasized once more that this review has only focussed on a single, albeit diverse, food category, DSs.It is understood that DSs make up only one aspect of a more varied dietary pattern with other foods that may contribute to the development of excess weight, in the wider, multifactorial obesogenic environment. 100The results discussed in this review should therefore be carefully considered with this in mind.

| Implications for practice and future research
The purpose of nutritional epidemiology is to provide a better understanding of how foods might affect the health of a particular population and furthermore to give insights into how dietary alterations might affect health outcomes.Although such dietary changes may be made by the individual, they are notoriously difficult to maintain longterm in our current food environment 101 ; and therefore, large-scale changes affecting the food environment itself, at a policy level, may be preferable. 102Extrapolation of the results of this systematic review is best interpreted in terms of specific populations and considering the study designs employed.In adults, while cross-sectional studies provided somewhat inconclusive results related to the effect of DS intake on weight status and food intake, a greater proportion of longitudinal studies highlighted that greater DS portion size and/or frequency of intake was associated with increased EI and weight status over time.Results from experimental studies, however, consistently demonstrate that a larger portion size or packaging size is associated with greater EI.As greater EI over time may contribute to the development of excess weight gain, policies aimed at providing, smaller, individual portions of dietary snacks may play a part in reducing overall EI.
Results in children, however, are far less conclusive and tend to, counterintuitively, suggest that greater DS portion size and/or frequency of intake was associated with reduced weight status.This observation requires further investigation to determine the role DS intake may play in childhood and adolescent obesity. 103This might take the form of investigations assessing the impact of different-sized DS portions on acute (<24 h) EI and how this may impact behavior and translate to longer term EI, body composition, and DQ.Further large-scale epidemiological studies that explore the influence of DS intake on DQ and how that may cluster with other health behaviors are warranted to promote long-term health.

| CONCLUSION
Increased DS consumption may contribute to EI in the diet and contribute to higher body mass levels as shown by experimental and longitudinal data.However, cross-sectional data, which are inadequate to determine clear conclusions, do not show consistent associations between DS intake and increased body mass/BMI.Given that experimental and longitudinal findings suggest that reducing the size of DSs could lead to decreased consumption and subsequent lower EI and body mass, food policymakers and manufacturers may find it valuable to consider altering the portion and/or packaging size of DSs.To support strategies for weight management, such policy changes may help consumers by altering product packaging, for example, providing smaller, individual-sized portions of DSs.Such a strategy may influence the quantity of DSs consumed.Currently, there is a lack of studies that have investigated a potential relationship between DS intake and DQ, which provides scope for further future studies.Furthermore, the negative associations of DS intake with weight status in children and adolescents require further investigation.
Abbreviations: BMI, body mass index; DS, discretionary snack; EI, energy intake; FFQ, food frequency questionnaire; FMI, fat mass index; T, time; NHLBI, National Heart Lung and Blood Institute; NI, no information; NIEHS, National Institute of Environmental Health; NIH, National Institutes of Health; NW, normal weight; OWOB, overweight/obesity; SD, standard deviation; SEP, socio-economic profile.

25 = 3
Generalized linear regression analysis kcal/d a (mean ± SD) BMI kg/m 2 b (mean ± SD), Association of frequency of DS consumption with total EI a O'Connor et al 2015 62 Mann-Whitney U-test Median (IQR) (g/10 MJ/day): Association of frequency of DS intake with BMI status NI BMI < 25 = 2.6 (0.6, 10.3) BMI > et al 2011 57 Mean (±SD) daily per capita intake (g) Regression analysis Food energy, kcal a (mean ± SE) BMI kg/m 2b mean ± SE Comparison of mean daily EI a USDA-Agricultural Research Service through a specific cooperative agreement.Partial support was received from the USDA Hatch Project.
investigated a potential relationship between chocolate consumption and markers of total and central body fat.Adolescents consuming the highest tertile of chocolate consumption had higher energy and saturated fat intake compared with those in the lower tertile but lower levels of central and total fatness, including BMI, regardless of relevant confounders.Using national data from American children (n = 11,181), O'Neill et al58 investigated the association between chocolate candy and sugar candy intake and body weight.Despite chocolate candy consumers having higher intakes of energy, total fat, saturated fatty acids, and added sugar, they had lower weight, lower waist circumference, and lower percentile BMI-for-age than non-consumers.Similarly, sugar candy consumers had lower weight, BMI waist circumference, and percentile/z-score for BMI-for-age than non-consumers.
investigated how the quality, quantity, and frequency of DS consumption differs in different BMI categories.Participants with a healthy weight showed a tendency to consume more DSs such as desserts, popcorn/pretzels/crackers, and candies/chocolate compared with participants living with excess weight.

%Children 4 daysMarchiori et al 2012 71 ANOVA
Body fat.Abbreviations: BMI, body mass index; DS, discretionary snack; EI, energy intake; FFQ, food frequency questionnaire; T, time; NHLBI, National Heart Lung and Blood Institute; NI, no information; NIH, National Institutes of Health; SD, standard deviation; SEP, socio-economic profile.T A B L E 5 Experimental studies reporting DSs and EI and/or weight status.Marchiori et al 2012 71 Effect of size of DS-on-DS intake a and EI b A: Mixed B: 9.2 ± 2.5 C: NI D: NI NA BMI 1 eating episode Adult Marchiori et al 2011 72 Effect of size of DS-on DS intake a and EI b A: Mixed B: 18-27 C: NI D: NI NA BMI 1 eating episode Haire and Raynor 2014 70 Association of package unit size with total DS consumed A: Mixed B: 23.7 +/À 3.3 years C: NI D: 81.3% white NA BMI Raynor and Wing 2007 73 Effects of package unit size on food and aAssociation between DS size and DS intake.bAssociation between DS size and Energy Intake.Abbreviations: BMI, body mass index; DS, discretionary snack; EI, energy intake; SD, standard deviation; SEP, socio-economic profile; NI, no information; NIH, National Institutes of Health.T A B L E 5 (Continued) Number of cookies consumed (mean ± SD) a EI (mean g/kcal) b Effect of size of DS-on-DS intake a and EI b reported intake of chocolate was positively associated, whereas sweets were negatively associated with living with overweight/obesity in children at 5 years of age.However, after adjustment for parental BMI, parental education, and heredity risk for DM, only frequency of intake of sweets was negatively associated with being classified as overweight or living with obesity.Phillips et al69 investigated the relationship between the consumption of energy-dense snack foods and relative weight change over 4 years in 196 adolescent girls living without obesity.There was no statistically significant relationship between total DS (cookies, cakes, pies, brownies, potato chips, corn chips, chocolate, and non-chocolate candy food) consumption, expressed as servings per day or as a percentage of daily calories, with body fat percentage.Over a longer time frame, O'Neil et al68 reported childhood (aged 10 years) candy consumption was not associated with weight status or body composition in young adulthood (aged 19-28 years).
These findings were similarly reported in children, in one study.Marchiori et al71 used a between-subjects randomized design to examine the influence of changing the size of snack food portions on short-term EI among Belgian children.Participants were offered the same weight of cookies, either full size or cut in two to make the portion size smaller.Decreasing the item size of food led to a decrease of 25% in gram weight intake, or a total of 68 fewer kilocalories consumed.
Cross-sectional studies reporting DSs and EI and/or weight status for children.NHLBI, National Heart Lung and Blood Institute; NI, no information; NIEHS, National Institute of Environmental Health; NIH, National Institutes of Health; NW, normal weight; OWOB, overweight/obesity; SD, standard deviation; SEP, socio-economic profile.
Abbreviations: BMI, body mass index; DS, discretionary snack; EI, energy intake; FFQ, food frequency questionnaire; FMI, fat mass index; T, time;T A B L E 2 (Continued) Cross-sectional studies reporting DSs and DQ.
a Association between DS size and DS intake.b Association between DS size and Energy Intake.Abbreviations: BMI, body mass index; DS, discretionary snack; EI, energy intake; SD, standard deviation; SEP, socio-economic profile; NI, no information; NIH, National Institutes of Health.T A B L E 6 Abbreviations: BMI, body mass index; DQ, diet quality; DS, discretionary snack; EI, energy intake; HEI, Healthy Eating Index; NI, no information; NIDDK, National Institute of Diabetes and Digestive and Kidney Diseases; NIH, National Institutes of Health; SD, standard deviation; SEP, socio-economic profile; USDA, United States Department of Agriculture.T A B L E 6 (Continued) Note: Letters (e.g., a, b ) indicated cross-referencing within studies for relevant analysis.Abbreviations: BMI, body mass index; DQ, diet quality; DS, discretionary snack; EI, energy intake; HEI, Healthy Eating Index; NI, no information; NIDDK, National Institute of Diabetes and Digestive and Kidney Diseases; NIH, National Institutes of Health; SD, standard deviation; SEP, socio-economic profile; USDA, United States Department of Agriculture.